Journal of International Economics 58 (2002) 335–358
www.elsevier.com / locate / econbase
Trade and foreign direct investment in China: a political
economy approach
Lee G. Branstetter a,c , Robert C. Feenstra b,c , *
b
a
School of Business, Columbia University, New York, NY 10027 USA
Department of Economics, University of California, Davis, CA 95616, USA
c
National Bureau of Economic Research, Cambridge, MA 02138 USA
Received 18 October 1999; received in revised form 5 February 2001; accepted 30 June 2001
Abstract
We view the political process in China as trading off the social benefits of increased trade
and foreign direct investment against the losses incurred by state-owned enterprises due to
such liberalization. A model drawing on Grossman and Helpman [Am. Econ. Rev. 84
(1994) 833; The Political Economy of Trade Policy: Papers in Honor of Jagdish Bhagwati,
MIT Press, Cambridge (1996) 199] is used to derive an empirically estimable government
objective function. The key structural parameters of this model are estimated using
province-level data on foreign direct investment and trade flows in China, over the years
1984–1995. We find that the weight applied to consumer welfare is between one-seventh
and one-quarter of the weight applied to the output of state-owned enterprises.
 2002 Elsevier Science B.V. All rights reserved.
Keywords: China; WTO; Foreign direct investment; Political economy
JEL classification: F21; H32
1. Introduction
In recent years, few developments in international economics have elicited more
interest or stirred more controversy than the sudden emergence of China as a
*Corresponding author. Tel.: 11-530-752-7022; fax: 11-530-752-9382.
E-mail address: [email protected] (R.C. Feenstra).
0022-1996 / 02 / $ – see front matter  2002 Elsevier Science B.V. All rights reserved.
PII: S0022-1996( 01 )00172-6
336 L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
trading nation and manufacturing center.1 China’s high rate of economic growth
since the adoption of more liberal economic policies under Deng Xiao-Ping in the
late 1970s, while perhaps exaggerated by the official statistics, is nevertheless a
stunning accomplishment.2 In the space of less than a generation, China has
transformed itself from a poor nation almost completely cut off from the global
economy to one of the world’s most important suppliers of labor-intensive
manufactures. Its arrival into the ranks of trading nations will be complete when it
has implemented the conditions for joining the World Trading Organization
(WTO).
Nevertheless, a number of economists have recently called into question the
sustainability of China’s current rates of GDP and export growth. These observers
have tended to point out the shortcomings of China’s socialist market economy,
focusing on the inability of the government to raise revenue by taxing the new
private economy, the problems the government has encountered in reforming state
owned enterprises, the distortions created by China’s dualistic trade regime, and
the mounting non-performing loans problem in the Chinese banking sector.3 The
reality of important distortions in the Chinese economy, and particularly in its
international trade and foreign direct investment regime, is increasingly evident.
We offer, in this paper, a theoretical and empirical application of the political
economy models pioneered by Grossman and Helpman (1994) to analyze these
distortions. We view the political process in China as trading off the social benefits
of increased trade and foreign direct investment against the losses incurred by
state-owned enterprises due to such liberalization. We write down an explicit
theoretical model of the policy formulation process which, we argue, captures
important features of Chinese institutional reality. We then use this model to derive
an empirically estimable governmental objective function. The key structural
parameters of this model are estimated using province-level data on foreign direct
investment and trade flows in China in the years 1984–1995.
This paper falls within a growing empirical analysis of endogenous political
economy, based on a Grossman–Helpman style political economy model. The first
paper of which we are aware that attempts this kind of analysis is Goldberg and
Maggi (1999), and there are now several more, including Gawande and
Bandyopadhyay (2000), Grether et al. (2002), Mitra et al. (2002), and McCalman
(2000). Our paper differs from these in focusing on FDI rather than tariffs as the
policy variable used to ‘reveal’ political preferences. In addition, the Goldberg and
Maggi paper and the work of Gawande and Bandyopadhyay are set in a very
1
For an excellent overview of this emergence, see Naughton (1996).
See Sachs and Woo (1997) for a discussion with some reference to the problems of official Chinese
statistics.
3
Again, Naughton (1996) provides an excellent overview of these issues.
2
L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
337
different institutional context—the United States. McCalman examines Australia,
while Grether et al. and Mitra et al. examine Mexico and Turkey, two developing
countries which underwent well-documented trade and investment liberalizations
over the course of the 1980s. In contrast, our paper explicitly models key features
of China’s trade and FDI regime—a regime which remains much less liberal
despite two decades of reform.
Our empirical estimates allow us to quantify the relative weights placed on
‘special interests’ and general economic welfare. In striking contrast to Goldberg
and Maggi’s (1999) results with US data, we find quantitative evidence that
Chinese provincial and central governments have given significantly greater
weight to the special interests of state-owned enterprises than to consumer
welfare—between four and seven times greater weight in our point estimates. Our
framework allows us to estimate the impact on governmental utility of proposed
changes in China’s trade regime. As an experiment, we estimate the impact of the
tariff reductions China has promised under its plan to join the WTO. The
implications of these findings for the political economy literature and for the
debate over economic reform in China are discussed in the conclusion. Additional
details on the model and data sources are described in our working paper
(Branstetter and Feenstra, 1999).
2. Foreign direct investment in China
A comprehensive description of Chinese economic reform, even one focused
solely on the evolution of China’s FDI regime, is well beyond the scope of this
paper. The reader is directed to Lardy (1992, 1998) and Naughton (1995) for three
well-regarded studies.4 Here, we quickly summarize some of the most important
policies related to FDI, and introduce the data that will be used in later estimation.5
The first column in Table 1 illustrates the cumulative stock of FDI into China’s
30 provinces (not counting the Hong Kong Special Administrative Region) in
1995, and this is graphed in Fig. 1. The variation across regions is quite striking,
with Guangdong province and neighboring Fujian together maintaining a dominant
position as the most important site of FDI activity. These two provinces are the
sites of the ‘special economic zones’ (SEZ), established in 1979 and giving
preferential tax and administrative treatment to foreign firms locating there.6 These
zones charged a reduced tax of 15% on business income of foreign affiliated firms
4
The material in this section also draws on Grub and Lin (1991) and Pomfret (1991).
The sources for data in Table 1 are described in a data Appendix available on request.
6
These SEZs consisted of three in Guangdong: Shenzhen (across the border from Hong Kong),
Zhuhai (across the border from Macau), Shantou (on the Guangdong coast facing Taiwan), and also
Xiamen in Fujiuan (directly across the Taiwan Straits from Taiwan).
5
338 L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
Table 1
Selected data in 1995
FDI
($ bill)
Multinational
share
State-owned
share
Import
share
Wage
premium
Tariffs
5.30
3.75
0.215
0.240
0.555
0.284
0.075
0.043
0.366
0.522
0.413
0.306
Coastal provinces
Liaoning
Hebei
Shandong
Jiangsu
Shanghai
Zhejiang
Fujian
Guangdong
Guangxi
Hainan
Average
5.38
2.02
8.51
14.01
11.61
4.01
13.73
39.02
2.79
4.00
10.51
0.042
0.066
0.054
0.102
0.290
0.075
0.270
0.271
0.065
0.204
0.144
0.389
0.327
0.274
0.176
0.294
0.082
0.068
0.000
0.357
0.054
0.202
0.022
0.007
0.007
0.008
0.080
0.010
0.035
0.075
0.014
0.348
0.061
0.475
0.36
0.308
0.353
0.452
0.27
0.387
0.33
0.357
0.436
0.373
0.227
0.289
0.282
0.223
0.163
0.240
0.298
0.215
0.252
0.172
0.236
Inland provinces
Heilongjiang
Jilin
Inner Mongolia
Shanxi
Henan
Anhui
Hubei
Jiangxi
Hunan
Guizhou
Yunnan
Sichuan
Tibet
Qinghai
Shaanxi
Gansu
Ningxia
Xinjiang
Average
1.26
1.07
0
0.24
1.53
1.2
2.1
0.91
1.46
0.21
0.57
2.35
0
0.02
1.25
0.22
0.1
0.23
0.82
0.026
0.062
0.048
0.017
0.041
0.032
0.037
0.041
0.034
0.032
0.028
0.034
0.000
0.006
0.045
0.040
0.093
0.020
0.035
0.648
0.576
0.606
0.433
0.323
0.292
0.354
0.495
0.385
0.641
0.644
0.374
0.725
0.827
0.563
0.645
0.665
0.715
0.551
0.011
0.054
0.021
0.005
0.006
0.005
0.013
0.008
0.009
0.015
0.038
0.012
0.250
0.005
0.018
0.013
0.006
0.009
0.028
0.582
0.367
0.232
0.399
0.352
0.578
0.35
0.468
0.437
0.43
0.27
0.453
0.621
0.463
0.533
0.396
0.352
0.245
0.418
0.264
0.174
0.320
0.216
0.294
0.241
0.195
0.257
0.209
0.139
0.234
0.186
0.321
0.157
0.234
0.146
0.170
0.092
0.214
Beijing
Tianjin
FDI is the cumulated stock of foreign investment. The shares of provincial spending on multinationals, state-owned enterprises, imports and town and village enterprises (TVE) are measured relative to
provincial apparent consumption. The tariff variable equals a weighted average of the 1996 ad valorem
tariffs in each province.
L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
339
Fig. 1. FDI stock in China ($ billion).
(as compared to 33% for domestic firms), though these taxes were not levied
during the first 2 years of operation and charged at one-half of the full rate in the
third through fifth years.7 In addition, wholly foreign owned enterprises were
permitted for the first time.
It is noteworthy that none of the original special economic zones were
developed industrial centers in 1979. In fact, these zones were established outside
the state’s industrial centers to prevent ‘contamination’ of Chinese heavy industry
by outside influences. These ‘experiments’ in attracting foreign direct investment
were quite successful. In 1984, 14 additional areas known as ‘open coastal cities’,
were granted similar exemptions from taxes and administrative procedures in a bid
to attract FDI.8 These cities levied a tax rate of 24% on foreign affiliated firms, and
were granted local authority to approve foreign investment for projects under $30
million (now $50 million). All of the coastal provinces listed at the top of Table 1
included ‘open coastal cities’ or SEZs, and these average over $10 billion in
7
Naughton (1996, p. 301). The discrepancy between the corporate tax for foreign-affiliated and
domestic firms has led many Chinese firms to seek foreign partners, sometimes through setting up
corporations in Hong Kong. This so-called ‘round tripping’ accounts for some (unknown) fraction of
the foreign investment into China.
8
The entire island of Hainan was declared a special economic zone in 1988.
340 L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
foreign investment by 1995. Shanghai and neighboring Jiangsu have received very
substantial inflows starting from a very low base in the late 1980s, such that they
have collectively become the next-largest recipient area by 1995, after Guangdong
and Fujian.
The next major regulatory change in FDI came in 1986, with the implementation of the so-called ‘Twenty-two Regulations’. These changes represented a major
liberalization which applied throughout China. Foreign invested enterprises (FIE)
were made eligible for reduced tax rates regardless of location, and were given
increased managerial autonomy. The establishment of an FIE was still subject to
the approval of local and central agencies, and in practice, there continues to be a
considerable degree of local autonomy in regulating FIEs. Their presence in the
inland provinces still lags considerably behind the coast, with an average FDI
stock of less than $1 billion in 1995. The inland provinces with the highest levels
of FDI are those in the far north (Heilongjiang and Jilin), or that border the coastal
provinces (such as Henan, Anhui, Hubei and Hunan, as well as Sichuan which
contains the industrial city of Chongqing).
Much of the foreign investment, especially that in the SEZs, has been for the
purpose of ‘processing trade’ rather than ‘ordinary trade’. Under the former
activity, intermediate inputs are imported, incorporated into other products, and are
then exported again.9 While this processing activity is important, it is not the focus
of our paper. Rather, we will be interested in foreign investment for the purpose of
selling products in China. To measure this activity, we isolate the sales of foreign
firms that are made to the domestic market. Let Output jit denote the sales of
firm-type j in province i in year t, where j 5 d for domestic state-owned firms, and
j 5 m for multinational FIEs. In addition, let Export jit denote the exports of these
firms. We measure apparent consumption in province i and year t by Total
Output it 2 Exports it 1 Ordinary Imports it . Note that imports for processing are
excluded from provincial apparent consumption. Then the share of state-owned
firms ( j 5 d), multinational firms ( j 5 m) and imports ( j 5 f ) in apparent
consumption are given by:
Output jit 2 Exports jit
]]]]]]]]]]]]]
Total Output it 2 Exports it 1 Ordinary Imports it
s jit 5
Ordinary Imports it
]]]]]]]]]]]]]
Total Output it 2 Exports it 1 Ordinary Imports it
5
for j 5 d, m,
for j 5 f.
The multinational share of provincial apparent consumption is shown in the
second column of Table 1. It has a correlation of 0.67 with the FDI stock in the
first column. This high correlation reflects the fact that provinces with a large
9
Borrowing from the Chinese terminology, Naughton (1996) refers to the two trade regimes as the
‘export processing’ and ‘ordinary trade’ regimes, respectively, and he provides an excellent discussion
of them.
L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
341
amount of FDI for processing trade will also have foreign firms selling locally. But
it is also the case that provinces with high FDI have a smaller share of
state-owned firms. Thus, the correlation between the FDI stock and the stateowned share (shown in the third column) is 20.63, and between the multinational
and state-owned shares is 20.59. The most extreme example of this is Guangdong, which has the highest FDI stock, a multinational share of 27%, and
virtually no presence of state-owned firms. As we have already mentioned, this
province was chosen as a site for SEZ precisely because the foreign firms would
not compete with any local state-owned firms, as well as for its proximity to Hong
Kong, of course.
The reverse situation applies in the inland provinces, most of which have
multinational shares less than 5% and state-owned shares exceeding 30%. The
presence of state-owned firms in the inland is no accident: Mao Zedong selected
these as the sites for heavy industries in the 1960s, as a protection against possible
military invasion.10 Our thesis will be that the continuing presence of inefficient
state-owned industries creates a local political force that may resist the entry of
competing, foreign firms. In some industries, foreign-invested enterprises have
already obtained a dominant market position, such that the state-owned enterprises
have been confined to the less profitable end of the market.11 A number of studies
suggest that the Chinese government, both national and local, is acutely aware of
this competition, and has taken steps to impede the ability of foreign firms to
compete in the Chinese market (e.g. Rosen, 1998). For example, multinationals
regularly confront a nexus of restrictions on their operations, including export
requirements, localization requirements, requirements for technology transfer, and
restrictions on domestic market access. Provincial and even local governments also
regularly attempt to extract funds from foreign firms through both legal and illegal
surcharges and taxes. While on paper foreign firms get favorable tax and import
treatment, in practice it is clear that foreign firms are often operating with a
government-engineered disadvantage.
The reader may notice that the shares of multinationals, state-owned firms and
imports do not sum to unity in Table 1. This is because a large share of output in
most provinces is supplied by ‘town and village enterprises’ (TVE), which are not
classified as state-owned; a small amount of output now also comes from
privately-owned enterprises. The TVEs were initiated as part of the agricultural
reforms of 1979, with the Chinese leadership granting rural municipalities
considerable freedom to set up manufacturing enterprises whose output and
production decisions would not be subject to the dictates of central planning.
10
As described by Naughton (1988), the provinces selected for this (economically disastrous) ‘third
front’ construction included all of Sichuan, Yunnan, Guizhou, Gansu, Qinghai and Ningxia, a portion of
Shaanxi and the western regions of Henan, Hubei and Hunan.
11
This view was strongly confirmed from our interviews with expatriate managers conducted in
China in December, 1998.
342 L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
These enterprises were quite successful, and by the end of the 1980s, the TVE
share exceeded the state-owned enterprise (SOE) share in most provinces.
There is an interesting time-series pattern of growth in TVE activity, and
therefore in the state-owned share. This is illustrated in Fig. 2 for Hebei province,
which surrounds metropolitan Beijing and the industrial center of Tianjin. The
patterns in Hebei are fairly typical of what we find in other eastern provinces. One
sees a clear trend break in the early 1980s, as the rural reforms which gave birth to
the township and village enterprises have their effect. The downturn in SOE output
share is literally the mirror image of the increase in the TVE output share. This
does not principally reflect a ‘crowding out’ of SOE production but rather a
growth in light ‘rural’ industry, which produces commodities and serves customers
who were underserved by the planned economy.12 The mid-1980s are marked by
an impressive rise in TVE output share, which comes to a halt at precisely the time
a ‘retrenchment’ in economic policy, following the protests at Tiananmen Square,
was phased in. During this period, economic policymakers deliberately steered
capital to the SOEs and deliberately set out to cut off access to capital for
non-state enterprises. This period of retrenchment was not sustained. In the 1990s,
Fig. 2. Changes in industrial output in Hebei Province, 1980–1995.
12
As Naughton (1995) has suggested, TVE activity was actually ‘suburban’ rather than ‘rural’ in
character. It has concentrated in rural zones near large industrial centers, such as Hebei Province (which
envelops Beijing and the industrial center of Tianjin) and the provinces of Jiangsu and Zhejiang (which
border metropolitan Shanghai). This geographic distribution suggests that TVEs were complementary
to SOEs, rather than being direct competitors.
L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
343
while there is some continued rise in the TVEs’ output share, the continuing
decline in SOE output share now largely reflects a striking rise in the output share
of proprietorships. It is also in this period that the FIEs’ output share increases the
most—but it remains small relative to the output shares of the other major
categories of non-state producers. The bottom line of this exercise is that the
policy record and our data strongly suggest that there is significant time-series
variation in the output of state-owned enterprises relative to total regional
industrial activity (in addition to the cross-sectional variation), and this is what we
will ultimately rely on in the estimation.
We conclude this section with a note on the level of intranational, as opposed to
international, economic integration in China, which is the subject of much current
debate among experts on the Chinese economy. On the one hand, Young (2000)
has argued on the basis of data on provincial production structure that economic
integration among Chinese provinces is limited and declining—that China has
become, ‘a fragmented internal market with fiefdoms controlled by local officials’.
At the other extreme, Naughton (1999) has argued that data on inter-provincial
flows of goods, taken from the input–output tables, suggests that the level of
interprovincial trade is relatively high and increasing. For reasons of modeling
convenience, we shall take a position similar to that of Young, although our results
will go through under the weaker assumption that interprovincial trade in goods
manufactured by FIEs (and the competing SOE-manufactured goods) is limited.
This view received surprisingly strong support from discussions with expatriate
managers and native Chinese executives. While the extent of local protectionism
varies across industries, it is seen as a barrier to growth and economic efficiency
by nearly every firm we interviewed. As one Shanghai-based advisor to American
firms put it, ‘There is no China market. There are several China markets, and none
of them are as big as you think’.13
3. The model
The model we develop draws heavily from Grossman and Helpman (1996), who
apply the political-economy framework developed in their 1994 paper to the issue
of multinationals. Specifically, they investigate whether the entry of multinationals
will affect the level of protection, and conversely, in a model where the domestic
industry is making campaign contributions. In our model, we will suppose instead
13
Naughton (1999) cites evidence that local governments have used discrimination in legal
proceedings, allocation of bank credit, and regulatory certification to prevent ‘outside’ firms based in
other provinces from succeeding in local markets. Another avenue of control has been local influence
over state-run wholesalers, who, until recently, played a large role in many commodity flows. Local
governments were able to block access of ‘outside’ goods to distribution channels.
344 L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
that the domestic industry is owned by the government, which gives extra weight
to industry profits in its objective function.14 The presence of multinationals
presents a potential threat to the state-owned enterprises through product market
competition. Li and Chen (1998) use a framework like this to examine the
government’s ability to extract rents from the multinationals, and we will use
policy instruments similar to theirs.
Because some features of our model are familiar, we will try to present it as
concisely as possible. The following points summarize the main components.
Regions: we will treat the regions within the country as distinct, following
Young’s (2000) thesis that the provinces within China have only limited trade
with each other. Thus, the model below describes each region, which will differ
in terms of their products, the number of domestic (state-owned) firms found in
each, and in the size of their labor force L.15
Products: each regional economy produces a numeraire good denoted by x 0 ,
and a differentiated product which is a CES aggregate of the various varieties.
There are three sources for the products: n d varieties are produced by domestic
firms; nf varieties are produced by foreign firms, where m of these are
produced locally by multinationals, and the remaining (nf 2 m) are imported
into the country. We will be treating n d and nf as exogenous variables, whereas
the number of multinationals m is the key endogenous variable.
Consumers: consumer preferences are given by
S
D
u
U 5 x 0 1 ]] x (u 21 ) / u , u . 0, u ± 1,
u 21
(1)
where the CES aggregate is,
x 5fn d x (d´ 21) / ´ 1 (nf 2 m)x (f ´ 21) / ´ 1 mx (m´ 21) / ´g ´ / (´ 21 ) , ´ . 1.
(2)
The values x j denote consumption of the differentiated varieties from source
j 5 d (domestic firms), f (imported), and m (multinationals). The parameter ´ is
the elasticity of demand for individual varieties, while u is the elasticity of
demand for the aggregate good x; we add the standard restriction that ´ . u.
Factors: we will suppose that labor is the only factor of production, and one
14
As Naughton (1996) and several other authors have noted, the Chinese government still relies on
remittances from state-owned enterprises for about two-thirds of its revenue. This is also true at the
provincial level. The provincial government of Yunnan is rumored to obtain nearly all of its revenue
from the provincial tobacco monopoly.
15
Since our model focuses on the tradeoff between state-owned enterprises and FIEs, we will abstract
from consideration of non-state domestic enterprises. FIEs pose a competitive threat to the relatively
capital-intensive SOEs, while the labor-intensive inexpensive manufactures of the TVE sector (and
proprietorships) tend to have limited overlap with the set of goods FIEs market in China for domestic
consumption.
L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
345
unit of the numeraire is produced with one unit of labor, so wages are unity. We
shall assume, however, that the multinational firms pay a wage premium of
(w 2 1) . 0.16 The wage premium is meant to proxy for a wide array of
possible benefits that multinationals bring, but which are not captured in our
model.
Firms: costs for the locally produced products are c j , j 5 d, m. Costs for
products produced abroad are cf , but also face a specific tariff of t, so that
marginal costs become cf 1 t. Prices are a standard markup over these marginal
costs, using the elasticity ´. We make the key assumption that:
c m , cf 1 t.
(3)
This assumption will ensure that prices charged by multinationals, pm , are less
than those for imports, pf . It follows that as more multinationals enter (m
increases) and these products are sold for pm , pf , the demand for products of
the domestic firms (x d ) will decline due to substitution away from these
products. This is how product market competition between the multinationals
and domestic firms is captured in the model.
Policy towards multinationals: each foreign firm faces the decision of whether
to supply locally through imports, or through setting up a local plant which
requires a fixed cost of F . 0. We suppose that the government also charges the
multinational a profit tax of l $ 0. This instrument is supposed to reflect the
vast range of actual policies used in China to extract rents from multinationals,
and not just the corporate tax on multinationals.17 By modeling these policies as
a tax on profits, we are abstracting from the inefficiencies caused by actual
policies (such as local content restrictions, for example).
The multinational net profits earned locally are thus (1 2 l)pm 2 F, where
pm 5 pm xm /´ is profits under the standard CES markup-pricing rule. Alternately, the multinational could just export from the home country, and earn
pf 5 pf xf /´. Thus, entry will occur if and only if (1 2 l)pm 2 F $ pf . This
condition is written as:
pf xf
pm x m
(1 2 l)]] 2 F $ ].
´
´
(4)
16
The existence of this wage premium is strongly confirmed in our data and in our interviews with
multinational managers.
17
In fact, foreign firms are often taxed at zero or reduced rates for the first years of operation. Despite
this, there are many ways that local and national agencies extract rents from the multinationals. For
example, the fact that most multinationals have had to use local partners reflects an implicit tax on their
profits, which are shared with the partner; similarly, the land-use fees that are commonly charged
reduce the multinationals’ profits, as do conditions of technology transfer and export requirements. Of
course, bribes paid to allow multinationals to enter are another example of the profit tax. Wei (1998)
argues that corruption in China, which includes the need for ‘questionable payments’, acts as a
significant deterrent to foreign direct investment.
346 L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
We will assume that when m 5 l 5 0 then (4) holds as a strict inequality. This
means that for some positive l, entry of multinationals will occur.
Government objective function: we can define the government’s objective
function over the various interest groups, beginning with consumer / worker
utility U, which receives a weight of a. We will suppose that the domestic firms
are state-owned, so these profits accrue to the regional and national government. Profits of the domestic firms are n d pd 5 n d pd x d /´. We give revenue from
state-owned firms a weight b in the objective function. The government
extracts rents mlpm from the multinationals, and also collects tariff revenue of
t (nf 2 m)xf . These two sources of revenue are each given weights of unity. The
objective function for each region is then defined by
G(m, l, t ) ; a U 1 b n d pd 1 mlpm 1 t (nf 2 m)xf .
(5)
We do not suppose that the government directly controls entry of multinationals. Instead, entry is influenced through the tax rate l and the tariff rate t,
so the number of multinational firms is written as a function m( l, t ).
Multinationals react in the expected manner to changes in these policies: when
(4) holds as an equality, then dm / dl , 0 and dm / dt .0.
We are now in a position to set up and solve the governments’ problem. We
suppose that the central and the regional governments jointly determine the
rents appropriated from the multinationals in each region, taking as given the
import tariffs.18 While the tariff schedule is common across regions, the fact
that different outputs are produced means that the tariffs effectively differ
across regions. Denoting regions by the subscript i, we let Gi [m i ( li , ti ), li , ti ]
denote the objective function for each region. Then the profit taxes are chosen
to solve:
max
li
O G fm sl , t d, l , t g
i
i
i
i
i
i
(6)
i
subject to m i ( li , ti ) # nf , which is the maximum number of foreign firms
willing to enter any region. Our strong assumption in writing the objective
function as in (6) is that the tax rate li for one region is separable from that in
another region. This is an extreme version of Young’s (2000) thesis that the
provinces in China have only limited trade with each other. It follows that they
will not be in competition with one another for foreign investment, because
multinationals can only serve the market where they locate. While this case
does not literally apply in China, it is certainly true that foreign firms face
restrictions on their ability to market outside their immediate area. By assuming
18
The tariffs themselves are chosen by the central government, in a problem more complex than (6)
that would include export gains from WTO membership. The impact of changing tariffs is discussed in
Section 6.
L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
347
an extreme form of these restrictions, we greatly simplify the governmental
decision problem.
Differentiating (6), the first-order condition is that ≠Gi / ≠m i 5 2 m i pmi /
(≠m i / ≠li ). This means that the gain from attracting one more multinational
(≠Gi / ≠m i ) is just balanced against the fall in the revenue li m i pmi when the tax
rate is lowered to attract another multinational. The first-order condition for (6)
over the choice of li is used to obtain an estimating equation.19 We denote all
other variables in region i with that subscript, though for simplicity we suppose
this does not apply to the foreign price pf , or the number of foreign firms nf . We
also add a time subscript to all relevant variables. Then directly from the
first-order condition, we obtain the following equation for the share of spending
on products of multinationals in region i and year t:
S
D
S
D
w it 2 1
w it 2 1
s mit 5 2 b s dit 1 h ]] 1 a (´ 2 1) ]] (s dit 1 sf it )
w it
w it
SD S
D
ti
m it
2 ´ ] sf it 1 ]]] sf it
pf
nf 2 m it
H
F
S
(´ 2 1)lit pmit
a´
a (´ 2 1)2 w it 2 1
1 ]] 1 ]]]]]] 1 ]]] ]]
w it
(´ 2 u )
(´ 2 u )(Fit 1 lit pmit )
(´ 2 u )
pft
ti
2 ´ ] ]]]]
pf
(Fit 1 lit pmit )
S DG
J
D
(7)
where s jit , j 5 d, f, m denotes the share of provincial apparent consumption on
domestic, imported or multinational products. The first term on the right of (7)
is the share of provincial consumption accounted for by state-owned firms,
which enters with the coefficient 2 b. Thus, the weight on state-owned firms in
the regional objective function is simply obtained as the coefficient on their
share in the regression (7). A high weight on the state-owned firms indicates
that in regions where these firms are more prevalent, the share of multinational
firms will be correspondingly reduced.
The next term on the right of (7) is the wage premium paid by multinationals, which has the coefficient h ; a (´ 2 1)(u 2 1) /(´ 2 u ), which is of ambiguous sign. Following this is the wage premium times the share of spending on
state-owned firms plus imports. When the wage premium is higher, we expect
that regions would be more willing to accept multinationals, and this is
confirmed by having a positive coefficient a (´ 2 1) on that variable. The
estimate of ´ itself comes from the next term, which is the ad valorem tariff
rate times the share of spending on imports. This term reflects the loss in tariff
19
See our working paper for the derivation, where we also discuss the second-order conditions.
348 L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
revenue as multinationals enter (as in Brecher and Diaz-Alejandro, 1977), and
the coefficient is 2 ´. Thus, combined with the former coefficient we can
recover an estimate of a. The final term on the second line of (7) reflects the
number of multinationals times the share of imports. For simplicity we treat nf ,
which is the number of foreign firms wanting to export or invest in China, as
constant over regions and time, and estimate it as a coefficient.
The terms on the last two lines of (7) are rather complex expressions of the
profit earned by importing into or producing in China, entering themselves and
interacted with the wage premium and tariffs. If profits are low, these terms will
be correspondingly small. We have no way of measuring these terms, but it
seems likely that they will vary systematically across regions, and possibly also
over time. Thus, we will model the terms on the last two lines of (7) as:
H
F
S
D S DG
(´ 2 1)lit pmit
ti
a (´ 2 1)2 w it 2 1
]]]]]] 1 ]]] ]] 2 ´ ]
w it
pf
(´ 2 u )(Fit 1 lit pmit )
(´ 2 u )
5 gi 1 dt 1 u it ,
pft
]]]]
(Fit 1 lit pmit )
J
(8)
where gi and dt are province and year fixed-effects, and u it is a random error
assumed to be uncorrelated with the variables on the first line of (7). Gathering
all these terms into fixed-effects plus an error is a strong assumption, but we
have no other way of dealing with them. With this assumption, we will be able
to estimate (7) with two-stage least squares (TSLS) applied to panel data.
4. Data and estimation issues
The data used to estimate (7) was already introduced in Section 2 and Table 1,
and is a panel of 30 provinces or autonomous regions, over the years 1984–1995.20
The wage data was available for workers in multinationals, state-owned enterprises, and urban collectives, with that in multinationals usually being the
highest and that in urban collectives usually the lowest. The wage premium is
measured as the wages paid by multinationals minus that in urban collectives,
divided by that in multinationals.
Two other variables are needed in the estimation. As a proxy for the number of
multinational firms m it in a province in a given year, we use the cumulated stock
of foreign direct investment actually utilized (in billions of US$), shown in Table
20
Sources for data are described in our working paper and a data Appendix, available on request. The
principal sources are the Xinjiang Statistical Bureau (1998) and trade data from the Customs General
Administration in China.
L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
349
1.21 It is interacted with the share of ordinary imports sf it on the right of (7), and
the coefficient is then 1 /(nf 2 m it ). For simplicity, we suppose that nf is large
enough so that this coefficient can be treated as constant over time and regions,
and re-write it as 1 /nf when estimating (7). The tariff variable is computed for
province i as a weighted average of ad valorem tariffs, based on the provincial
imports of different goods multiplied by the 1996 tariff rate assessed on those
goods.22 Unfortunately, we did not have access to any earlier years of tariff data,
so the tariff rate ti /pf in (7) does not vary over time. All these variables are shown
in Table 1 for 1995.
Turning to estimation issues, the dependent variable in (7) is the share of
regional apparent consumption accounted for by products sold by multinationals,
while the state-owned and import shares appear on the right. Taken together with
the share of provincial spending on TVEs and individual proprietorships, these
shares would sum to unity, so there is some obvious endogeneity between the
multinational share on the right of (7), and the state-owned and import shares on
the left. In our working paper, we attempted to correct for this by estimating (7)
with instrumental variables, which were the same variables that appear on the right
of (7) but measured as levels rather than shares. Because the exogeneity of the
instruments could be questioned, here we take the approach of directly estimating
(7) in levels rather than shares, and using other instruments.
That is, we now let s jt , j 5 d, f, m denote the nominal provincial spending on
products of state-owned, imports or multinational firms, rather than shares. There
can still be an endogeneity problem, of course, if multinational firms ‘crowd out’
imports or state-owned firms. Other explanatory variables in (7) might also be
endogenous, such as the wage premia and import shares. Accordingly, we use
instruments for all the explanatory variables, and estimate (7) with TSLS.23 In the
21
This was computed by taking flows of foreign direct investment from the regional data and
cumulating them using a depreciation rate of 4.5% This depreciation rate was based on data drawn
from the 1994 Benchmark Survey on US direct investment abroad conducted by the Bureau of
Economic Analysis of the US Department of Commerce. We are grateful to Michael Ferrantino of the
International Trade Commission for drawing our attention to this document, and performing the
computation that produced this number.
22
We included in this calculation an indicator variable, equal to one when that product is also
exported by FIEs in the province, and zero otherwise. Thus, the weights used to compute the average
tariff were the provincial imports of different goods times the indicator variable, indicating whether the
good was also exported (and therefore produced) by FIEs.
23
The instruments are: provincial electricity production (used as an instrument for expenditure on
state-owned firms); indexes of urban, state and overall wages (used as instruments for the wage
premium); provincial GDP, population, average rural and urban income (used as instruments for
apparent consumption); provincial processing imports (used as an instrument for ordinary imports) and
various interactions between these terms.
350 L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
estimation we include provincial apparent consumption as a control variable, and
also experiment with using provincial GDP as a weight.
Another estimation issue is that the multinational share can be zero for some
provinces and years. This raises technical difficulties, however, because we also
have fixed-effects as in (8), and some of the right-hand side variables are
endogenous. There is no available estimator for a censored model with fixed
effects and endogenous variables. However our sample does not start until 1984,
and most provinces registered positive levels of FDI by or shortly after that date.
The only province which did not have any foreign investment by the end of our
sample period was Tibet, where political factors other than those we have modeled
are surely important. For these various reasons, we restrict the dependent variable
in (7) to positive values.24
5. Estimation results
Results from regressions based on (7)–(8), using data for 1984–1995, are
presented in Table 2. Each of the four columns of results uses a different
estimation technique, beginning with OLS, and moving to weighted TSLS.
Standard errors are placed in parentheses below the respective parameter estimates, and we see that our key parameters are estimated with varying degrees of
precision in these four columns. OLS estimates are largely uninformative, as we
are unable to obtain reasonable estimates of ´. However, our estimates improve
substantially when we move to two-stage least squares. The coefficients are
affected somewhat as we weight by provincial GDP in the third and fourth
regressions. Each regression includes a full set of provincial and year fixed effects.
An estimate of the parameter b, the weight of state-owned enterprise output in
the government’s objective function, can be taken from the regression coefficient
on state-owned enterprise output, and ranges from 1.3 to 1.66 (ignoring the OLS
estimate).25 Estimates of a, the weight on consumer welfare, can be derived from
the regression coefficients shown in the second and third rows of Table 2. For our
preferred specification, which is the weighted TSLS estimate in the third
regression, we obtain aˆ 5 0.24 as reported in Table 3. The difference between the
state-owned and consumer weight is bˆ 2 aˆ 5 1.42, and the standard error of this
difference is computed as 0.46. Thus, we find that the weight given to consumer
welfare is significantly lower than that applied to the output of state-owned
enterprises (at the 5% level), with their ratio being about one-seventh in our
24
This restriction only impacts 11% of our observations.
We note that the sign and magnitude of this estimate is contingent on including total provincial
consumption as a control variable in the regression: controlling for total provincial spending, a decline
in the spending on state-owned firms is associated with a rise in the spending on multinationals; but
without this control variable, the sign of bˆ is reversed.
25
L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
351
Table 2
Dependent variable—provincial spending on output of multinational enterprises
Coefficient
Independent variable:
Provincial spending
on state-owned
Wage premium
3(state1imports)
Tariff3imports
Estimation method
2b
a (´ 2 1)
2´
Wage premium
a (´ 2 1)(u 2 1)/(´ 2 u )
FDI stock
3imports
Provincial apparent
consumption
Provincial spending
on collectives
R 2, N
1/nf
Fixed effects
OLS
TSLS
Weighted
TSLS
Weighted
TSLS
20.41
(0.12)
0.42
(0.23)
20.15
(2.7)
2118
(64)
0.037
(0.022)
0.11
(0.02)
21.30
(0.40)
2.19
(0.87)
27.4
(3.3)
2967
(435)
0.037
(0.017)
0.15
(0.03)
21.66
(0.48)
3.62
(1.16)
215.9
(5.1)
23333
(1085)
0.052
(0.013)
0.15
(0.02)
0.85,
297
Yes
0.74,
280
Yes
0.76,
280
Yes
21.60
(0.45)
3.45
(1.16)
216.5
(4.8)
23430
(1130)
0.057
(0.012)
0.22
(0.07)
20.12
(0.10)
0.75,
280
Yes
Notes: the sample consists of 29 provinces (excluding Tibet) over 1984–1995, using fixed effects for provinces and for time; only
provinces with positive multinational output are included. All regressions except the first are estimated with TSLS. The weighted
regressions use provincial GDP as a weight. White standard errors are reported in parentheses.
preferred estimates. Without using provincial GDP as weights, the TSLS estimates
in the second regression gives aˆ 5 0.34, which is still one-quarter of the weight
given to the state-owned enterprises.
Turning to other parameters, we obtain high estimates of the elasticity of
substitution ´ (from 7 to 16), though these are somewhat imprecise. In the
Table 3
Coefficient estimates, by time period
Coefficient
Sample
1984–1995
(N 5 280)
1988–1995
(N 5 210)
1990–1995
(N 5 132)
a
b
´
0.24
(0.07)
0.24
(0.13)
0.20
(0.15)
1.66
(0.48)
1.32
(0.50)
1.04
(0.49)
15.9
(5.1)
11.5
(6.7)
10.0
(7.6)
Notes: computed from the third regression reported in Table 2, but run over different samples. The
estimation method is weighted TSLS. White standard errors are in parentheses.
352 L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
regression shown in the last column of Table 2, we experiment with including the
provincial spending on products from collectives (TVEs) as an additional
regressor. The idea here is that if collectives command the same political power as
state-owned enterprises (which we find unlikely), then the coefficient on provincial
spending on collectives should be comparable to that on the state-owned
industries. In fact, the coefficient on collectives is statistically insignificant, and its
inclusion has little impact on the other coefficients.
A breakdown of the implied structural coefficients of interest is provided in
Table 3 for later sub-samples of the data, the 1988–1995 period and the
1990–1995 period. As we confine our view to the later sub-samples, we lose
observations and consequently precision in some of our estimates; sub-periods
smaller than 1990–1995 do not yield many significant coefficients at all. In the
later sub-periods, the estimated magnitudes of the weight on state-owned enterprises falls. This is consistent with the historical trend towards liberalization, of
course. In the relatively liberal 1990–1995 sub-period, for example, we find that
the weight on state-owned enterprises is unity, which is still considerably higher
than the weight on consumer welfare (though the difference between these
estimates is no longer significantly different from zero). These point estimates are
consistent with the pessimism among some China scholars concerning the
momentum of reform of restrictions on FDI and trade.
As a final evaluation of our model and estimates, we see how well they can
reproduce correlations among key variables. Of special interest is the crossprovincial correlation between the multinational and state-owned shares, which is
20.59 in 1995. This negative relationship led to our hypothesis of product-market
competition between those firms: provinces with more state-owned firms appear to
be less willing to accept foreign firms. At first glance, our estimate of bˆ 5 1.66
might appear ‘too high’ to be consistent with this correlation. To check this, we
report in Table 4 various sample and predicted correlations between the multinational share, s mit , and other variables for 1984–1995. Over the entire sample, s mit
Table 4
Correlations with sample and predicted provincial share of spending on multinationals, 1984–1995
Correlation with:
Multinational share
FDI
State-owned share
Wage premium
Tariffs
Sample
Correlation with predicted share
correlation
b 5 1.7, ´ 5 16
1.7, 13
1.0, 8
0.50, 4.5
1.00
0.62
20.58
0.35
0.13
0.27
0.22
20.41
0.91
0.00
0.37
0.33
20.59
0.87
0.05
0.37
0.36
20.60
0.86
0.05
0.38
0.41
20.60
0.84
0.05
Notes: the predicted provincial share of spending on multinationals is computed using the formula
s mit 5 2 b s dit 1 a (´ 2 1)[(w it 2 1) /w it ](s dit 1 sf it ) 2 ´(ti /pft )sf it 1 (1 /nf )m it sf it , with parameters a 5
0.25, 1 /nf 5 0.05, and b and ´ as shown above. All correlations have N 5 297.
L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
353
has a correlation of 0.62 with provincial FDI, 20.58 with the provincial share of
spending on state-owned firms, 0.35 with provincial wage premia, and 0.13 with
provincial tariffs. How well do these correlations match up with our estimates?
Obviously, the fixed-effects included in (8) will raise the panel correlations, so let
us ignore the fixed effects. We then compute the predicted spending on multinationals, sˆ mit , using our parameter estimates for 1984–1995 and the terms on the
first line of (7). Because the coefficient hˆ on the wage premium is estimated with
such uncertainty, we omit this variable. The resulting correlations between sˆ mit and
other variables are shown in Table 4.
Using our TSLS estimates of b 5 1.7 and ´ 5 16, the correlation between the
predicted multinationals share sˆ mit and the other variables do not correspond very
closely to the sample correlations. However, a modest change in the elasticity of
substitution ´, from 16 to 13, leads to a much closer match for the correlation
between multinational and state-owned firms. Thus, the predicted multinational
shares sˆ mit have a correlation of 20.59 with actual state-owned shares, nearly the
same as the sample correlation. The predicted correlation with FDI is still too low,
while with the wage premium is too high (this may be due to the fact that we have
left the wage premia out of the formula for sˆ mit due to uncertainty about its
coefficient). It turns out that as we reduce the values of ( b, ´), from (1.7, 16) to
(1.0, 8) and then (0.5, 4.5) or even lower, the predicted correlations change very
little. That is, we still obtain a good match between the sample and predicted
state–multinational correlation, but this type of exercise does not identify the
weight b given to state-owned firms: it could take on a whole range of values,
depending on ´. We conclude that while our model can quite easily predict the
cross-provincial correlation of state-owned and multinational firms, identification
of b requires the full structural estimation, such as we have performed above.
6. Choosing the tariffs
At the beginning of the paper, we motivated our topic by the pending entry of
China into the WTO, and suggested that the existing trade barriers would have to
be reduced substantially to meet these commitments. To what extent does the
presence of multinationals interact with that goal? The willingness of foreign firms
to invest in China certainly reflects, in part, the difficulty of importing there
directly. If tariffs are reduced substantially, this raises the distinct possibility that
foreign firms who are not just involved in processing trade will divest of some of
their holdings. This scenario was explicitly mentioned in some of our field
interviews, and reflects the fact that for firms not involved in processing trade,
investment in China is an attempt to access the huge domestic market; low wages
play only a secondary role.
If we accept this characterization of foreign direct investment, it suggests that a
reduction in tariffs may lead to an outflow of foreign direct investment. There is no
354 L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
‘natural experiment’ that allows us to directly estimate this from data for China.
However, we can use the estimates from the previous section to construct the
outflow that is consistent with our theoretical model, in an attempt to judge the
welfare impact. To this end, we compute the change in the objective function of a
single region Gi [m i ( l *i , t ), l *i , ti ], when the multinational profit tax l *i has been
chosen optimally, as:
dG
≠G
≠G ≠m
]i 5 ]i 1 ]i ]i
dti
≠ti
≠m i ≠ti
≠Gi
≠m i / ≠ti ≠Gi
dli
5 ] 2 m i pmi ]]] 5 ] 1 m i pmi ]
≠ti
≠m i / ≠li
≠ti
dt
U
mi
.
(9)
The first line of (9) uses the envelope theorem, whereby we can hold l *i
constant when computing the effect on Gi of changing t. The second line makes
use of the first-order condition (7), and then the third line follows by computing
dli / dt u m i . 0 as the slope of an iso-curve of the function m i ( li , ti ). Thus, the total
impact of decreasing tariffs can be interpreted as: (i) the direct impact holding the
profit tax and number of multinationals fixed; and (ii) the indirect effect through
decreasing the profit tax itself so that multinationals do not exit the country. In
other words, rather than computing the outflow of multinationals due to a tariff
cut, we can (equivalently) compute in (ii) the drop in the profits tax imposed on
the multinationals so that they do not leave following a tariff cut.
To determine the relevant tariff cut, we note that the unweighted average of the
1996 tariffs in China is 23.4%, and under its WTO entry plan the tariffs in 2005
would be reduced to 16.2%, or three-quarters of their 1996 value.26 Applying this
reduction to the 1996 tariffs used in our estimation, shown in the last column of
Table 1, we obtain the proposed WTO tariffs shown in the third column of Table
5, at the provincial level. This drop in tariffs is used to evaluate effects (i) and (ii),
computed using the functional forms of our model and the previously estimated
parameters, with results shown in Table 5.
We show results in Table 5 for two values of the elasticities: low values of (´,
u ) 5 (5, 0.5), and high values of (´, u ) 5 (15, 2). We experiment with the weights
(a, b ) 5 (0.25, 1.7), as shown for the full sample in Table 3, or (a, b ) 5 (0.2, 1)
as shown for the 1990–1995 period; the value 1 /nf 5 0.05 is also used. The direct
effect of lowering the tariffs, which depends on how the current level compares to
the ‘optimal’ tariff, and the indirect effect through lowering the profit tax for
multinationals, are both measured relative to import expenditure. For brevity, we
report only the welfare effects for Beijing, Tianjin, and the averages of the coastal
and inland provinces.
26
These unweighted tariffs are reported in World Bank (1997, Annex Table A.1). If we take the
longer period from 1992 to 2005, then unweighted tariffs are reduced by one-half. Thus, to evaluate
this total change, the welfare effects in Table 3 should all be doubled.
L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
355
Table 5
Welfare effects of tariff reductions (expressed as percent of import value)
Initial
WTO
Example 1 (´ 5 5, u 5 0.5)
Example 2 (´ 5 15, u 5 2)
tariffs
tariffs
Direct
effect
Indirect
effect
Total
effect
Direct
effect
Indirect
effect
Total
effect
Weights (a, b ) 5 (0.25, 1.7)
Beijing
0.413 0.310
Tianjin
0.306 0.230
Coastal average 0.236 0.177
Inland average
0.214 0.170
21.9
20.7
20.7
24.1
24.8
23.3
23.5
20.4
26.7
24.1
24.2
24.5
32.8
17.4
11.0
6.9
24.4
23.0
23.4
20.4
28.5
14.4
7.6
6.5
Weights (a, b ) 5 (0.2, 1)
Beijing
0.413
Tianjin
0.306
Coastal average 0.236
Inland average
0.214
2.2
0.7
20.1
22.2
24.8
23.3
23.5
20.4
22.6
22.6
23.6
22.6
36.5
19.0
11.7
8.5
24.4
23.0
23.4
20.4
32.2
16.0
8.3
8.2
0.310
0.230
0.177
0.170
Notes: other parameters used in both examples are 1 /nf 5 0.05. The values shown give the welfare
effects of reducing tariffs from their ‘initial’ to ‘WTO’ levels, measured as a percentage of import
expenditure. The direct effect is calculated as the first term in (9), and the indirect effect as the second
term.
In the first case, with (a, b ) 5 (0.25, 1.7) and low elasticities, Beijing has a
direct welfare loss of nearly 2% of its import expenditure, reflecting the fact that
even its high initial tariff (41%) is below the ‘optimum’ given the high weight on
state-owned enterprises and the low elasticities. Beijing also suffers a loss of
nearly 5% of imports from lowering the rents extracted from multinational firms,
for a total loss of nearly 7%. Losses of a smaller magnitude are found for the
neighboring industrial region of Tainjin, and for the coastal and inland provinces
on average. If instead we consider the weights (a, b ) 5 (0.2, 1), then the direct
losses turn to gains for Beijing and Tianjin, since their initial tariffs are above the
‘optimum’. Nevertheless, the indirect losses created by the potential exit of
multinational firms still leads to overall losses for these areas and the other
provinces. Thus, this example with low elasticities gives us a pessimistic scenario,
with losses due to trade liberalization. Of course, we have not factored into this
calculation the gains due to tariff removal abroad, which are essential to the
Chinese interest in WTO membership. Nevertheless, press reports from China at
the time of the WTO deal emphasized that, at the provincial level, local officials
and managers were focused more on the competition created with state-owned
firms, especially in the inland provinces.
In the second example we consider higher elasticities, (´, u ) 5 (15, 2), in which
case the ‘optimal’ tariffs are lower. Regardless of which values for (a, b ) are
used, we find that nearly all regions now have direct gains from tariff removal,
which are substantially larger than the indirect losses. For example, Beijing has a
direct gain of 33–37% of its import value, which is much larger than the 4%
indirect loss due to potential exit of multinationals. Gains of a smaller magnitude
356 L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
are observed for most other provinces. Thus, with high values for the elasticities,
most provinces gain from the proposed reduction of tariffs under China’s WTO
commitments, but for low values of the elasticities this is not the case. Both these
cases are within the range of elasticities that we have estimated. So while we are
not able to make more definite conclusions about the potential losses due to
multinational exit, and whether these offset the gains due to tariff removal, this
possibility is certainly within the range of our estimates.
7. Conclusions
After two decades of substantial reform, China’s current trade and foreign
investment regime remains far from open. If ever we needed a theoretical
framework to explain sharp deviations of policy away from that proscribed by
laissez faire economics, we need it here. Our paper represents a first attempt to
apply a theoretical model based on Grossman and Helpman (1994, 1996) to a
policy context in which political economy considerations are essential for
understanding and predicting the trajectory of economic reform, and then to obtain
economically meaningful estimates of the model’s parameters from Chinese
economic data.
As in any ambitious attempt to marry a complicated structural model with
imperfect data, we have been driven to make a number of compromises.
Nevertheless, the results that we end with are quite striking. To summarize, using
the estimates based on our full sample, we find that the weight applied to the
output of state-owned enterprises is much larger (four to seven times in our point
estimates) than the weight applied to consumer welfare. The evidence of a political
premium on state-owned production diminishes over time, but the point estimates
still indicate these firms are favored. Finally, we estimate the impact on regional
governmental objective functions of the proposed changes in tariff structure that
China has put forth as part of its bid for WTO accession. We find that these
changes could potentially lower the regional objective function, due to the exit of
multinationals. While this result is sensitive to the elasticities that are used, it
provides some quantitative backing for skepticism that China, given the current
political equilibrium, would actually follow through with the proposed liberalization.
On the surface, it might seem that our results are too pessimistic, given the fact
that a WTO deal—one which included surprising concessions on the Chinese
side—was successfully brokered at the end of 1999. However, we note that there
is a difference between signing a trade treaty and fully implementing its
provisions.27 Already, in the fall of 2000, there were press accounts of attempts by
27
Unfortunately, China’s past behavior gives one ample grounds for doubt on this point. For a
fascinating examination of China’s persistent failure to honor international copyright and trademark
laws, see Behar (2000).
L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
357
Chinese government officials to obviate or ignore promises of liberalization in key
sectors, such as telecommunications.28 Our interviews of expatriate managers in
China strongly indicate that these individuals believe tariff cuts will be at least
partially undone by the simultaneous construction of more subtle non-tariff
barriers, and the view that government will continue to favor ‘local production’ in
the post-WTO era is widespread.
As the focus of government development efforts shifts from the coast to the
Western provinces, it is already clear that one of the barriers will be the deeply
entrenched role of state-owned enterprises in this region of the country. Real
liberalization, market-oriented development, and successful attraction of foreign
direct investment would all bring about a decline in the state-owned sector of the
economy–and we can expect this to be resisted, especially in the inland provinces.
Unfortunately, we think that the political tradeoffs modeled in our framework—
and the somewhat pessimistic view of the political equilibrium implied by our
empirical results—may be quite relevant to the reality of the interior provinces for
some time to come.
Acknowledgements
We thank Chen Chun-Lai, Wen Hai, David Li, Barry Naughton, Daniel Rosen,
Scott Rozelle, Edward Steinfeld, William Skinner, Eric Thun, Shang-Jin Wei, Wing
Woo, Alwyn Young, seminar participants at the University of Michigan and the
World Bank, and three anonymous referees for helpful comments and suggestions.
We are indebted to a large team of research assistants, including Natasha Hsieh,
Songhua Lin, Kaoru Nabeshima, Shunli Yao, and David Yue. Branstetter wishes to
thank the Japan Foundation Center for Global Partnership and the Social Science
Research Council for financial support of the fieldwork component of this project,
and he thanks the many anonymous China-based multinational executives who
provided their insights concerning the environment confronting foreign firms in
China. He also acknowledges the support of a University of California Faculty
Research Grant. Feenstra thanks the National Science Foundation for financial
support. All errors are the sole responsibility of the authors.
References
Behar, R., 2000. Beijing’s phony war on fakes. Fortune 30 (Oct), 189–208.
Branstetter, L., Feenstra, R., 1999. Trade and foreign direct investment in China: a political economy
approach. NBER, Working paper no. 7100.
28
See, among many other contemporaneous press accounts, Silva (2000).
358 L.G. Branstetter, R.C. Feenstra / Journal of International Economics 58 (2002) 335–358
Brecher, R., Diaz-Alejandro, C., 1977. Tariffs, foreign capital and immiserizing growth. Journal of
International Economics 7, 317–322.
Gawande, K., Bandyopadhyay, U., 2000. Is protection for sale? A test of the Grossman–Helpman
theory of endogenous protection. Review of Economics and Statistics, 139–152, February.
Goldberg, P., Maggi, G., 1999. Protection for sale: an empirical investigation. American Economic
Review 89 (5), 1135–1155.
Grether, J.-M., de Melo, J., Olarrega, M., 2002. Who Determine Mexican Trade Policy? Journal of
Development Economics, forthcoming.
Grossman, G.M., Helpman, E., 1994. Protection for sale. American Economic Review 84, 833–850.
Grossman, G.M., Helpman, E., 1996. Foreign investment with endogenous protection. In: Feenstra,
R.C., Grossman, G.M., Irwin, D.A. (Eds.), The Political Economy of Trade Policy: Papers in Honor
of Jagdish Bhagwati. MIT Press, Cambridge, pp. 199–224.
Grub, P.D., Lin, J.H., 1991. Foreign Direct Investment in China. Quorum Books, New York.
Lardy, N., 1992. Foreign Trade and Economic Reform in China, 1978–1990. Cambridge University
Press, New York.
Lardy, N., 1998. China’s Unfinished Economic Revolution. Brookings Institution Press, Washington,
DC.
Li, D.D., Chen, Q., 1998. Why aren’t foreign direct investments always welcome?—A political
economy analysis. University of Michigan, Mimeo.
Mitra, D., Thomakos, D., Ulubasoglu, M., 2000. ‘‘Protection for Sale, In a Developing Country:
Democracy versus Dictatorship.’’ Review of Economics and Statistics, forthcoming
McCalman, P., 2000. Protection for sale and trade liberalization: an empirical investigation. Department
of Economics, University of California, Santa Cruz, Mimeo.
Naughton, B., 1988. The third front: defense industrialization in the Chinese interior. The China
Quarterly 115, 351–386, September.
Naughton, B., 1995. Growing Out of the Plan: Chinese Economic Reform, 1978–1993. Cambridge
University Press, New York.
Naughton, B., 1996. China’s emergence and prospects as a trading nation. Brookings Papers on
Economic Activity 2, 273–343.
Naughton, B., 1999. How much can regional integration do to unify China’s markets. UCSD, Mimeo.
Pomfret, R.W.T., 1991. Investing in China: Ten Years of the Open Door Policy. Iowa State University
Press, Ames.
Rosen, D., 1998. Behind the Open Door: Foreign Firms in China. Institute for International Economics,
Washington, DC.
Sachs, J.D., Woo, W.T., 1997. Understanding China’s economic performance. NBER, Working paper
no. 5935, February.
Silva, J., 2000. As trade bill is signed, China vows to live up to agreement. Crain’s Communications
19, 30.
Wei, S.-J., 1998. Why Does China Attract so Little Foreign Investment?. Harvard University and
NBER, Mimeo.
World Bank, 1997. China Engaged: Integration with the Global Economy. World Bank, Washington,
DC.
Xinjiang Statistical Bureau, 1998. China’s Regional Economy in the 17 Years Since Reform and
Liberalization Electronic Database on Diskette.
Young, A., 2000. The razor’s edge: distortions and incremental reform in the People’s Republic of
China. Quarterly Journal of Economics 115 (4), 1031–1036.